Noisy Speech Signal Research Articles

Spectral Reconstruction and Noise Model Estimation Based on a Masking Model for Noise Robust Speech Recognition

An effective way to increase noise robustness in automatic speech recognition (ASR) systems is feature enhancement based on an analytical distortion model that describes the effects of noise on the speech features. One of such distortion models that has been reported to achieve a good trade-off between accuracy and simplicity is the masking model. Under this model, speech distortion caused by environmental noise is seen as a spectral mask and, as a result, noisy speech features can be either reliable (speech is not masked by noise) or unreliable (speech is masked). In this paper, we present a detailed overview of this model and its applications to noise robust ASR. Firstly, using the masking model, we derive a spectral reconstruction technique aimed at enhancing the noisy speech features. Two problems must be solved in order to perform spectral reconstruction using the masking model: (1) mask estimation, i.e. determining the reliability of the noisy features, and (2) feature imputation, i.e. estimating speech for the unreliable features. Unlike missing data imputation techniques where the two problems are considered as independent, our technique jointly addresses them by exploiting a priori knowledge of the speech and noise sources in the form of a statistical model. Secondly, we propose an algorithm for estimating the noise model required by the feature enhancement technique. The proposed algorithm fits a Gaussian mixture model to the noise by iteratively maximising the likelihood of the noisy speech signal so that noise can be estimated even during speech-dominating frames. A comprehensive set of experiments carried out on the Aurora-2 and Aurora-4 databases shows that the proposed method achieves significant improvements over the baseline system and other similar missing data imputation techniques.

Research on Speech Endpoint Detection Algorithm with Low SNR

The detection of speech endpoint is an important application for speech signal processing. Although there are variance methods, the endpoint can’t be detected accurately in low SNR (Signal to Noise Ratio). The paper pointes out an endpoint detection algorithm combining two methods together: the one is improved spectral subtraction based on multitaper spectral estimation, and the other is BARK subband variance in frequency domain. Firstly, the noisy speech signal is processed though the improved spectral subtraction based on multitaper spectral estimation. It can achieve the purpose of noise reduction through this step. Then the noisy speech signal is detected using the method of BARK subband variance in frequency domain. Compared with the common endpoint detection algorithm, it is concluded that endpoint detection accuracy by new method can be improved in low SNR.

A Method for Voiced/Unvoiced Classification of Noisy Speech by Analyzing Time-Domain Features of Spectrogram Image

This paper presents a voiced/unvoiced classification algorithm of the noisy speech signal by analyzing two acoustic features of the speech signal. Short-time energy and short-time zero- crossing rates are one of the most distinguishable time domain features of a speech signal to classify its voiced activity into voiced/unvoiced segment. A new idea is developed where frame by frame processing has done in narrow band speech signal using spectrogram image. Two time domain features, short-time energy (STE) and short-time zero-crossing rate (ZCR) are used to classify its voiced/unvoiced parts. In the first stage, each frame of the analyzing spectrogram is divided into three separate sub bands and examines their short-time energy ratio pattern. Then an energy ratio pattern matching look up table is used to classify the voicing activity. However, this method successfully classifies patterns 1 through 4 but fails in the rest of the patterns in the look up table. Therefore, the rest of the patterns are confirmed in the second stage where frame wise short-time average zero- crossing rate is compared with a threshold value. In this study, the threshold value is calculated from the short-time average zero-crossing rate of White Gaussian Noise (wGn). The accuracy of the proposed method is evaluated using both male and female speech waveforms under different signal-to-noise ratios (SNRs). Experimental results show that the proposed method achieves better accuracy than the conventional methods in the literature.

Speech enhancement based on neural networks improves speech intelligibility in noise for cochlear implant users

Speech understanding in noisy environments is still one of the major challenges for cochlear implant (CI) users in everyday life. We evaluated a speech enhancement algorithm based on neural networks (NNSE) for improving speech intelligibility in noise for CI users. The algorithm decomposes the noisy speech signal into time-frequency units, extracts a set of auditory-inspired features and feeds them to the neural network to produce an estimation of which frequency channels contain more perceptually important information (higher signal-to-noise ratio, SNR). This estimate is used to attenuate noise-dominated and retain speech-dominated CI channels for electrical stimulation, as in traditional n-of-m CI coding strategies. The proposed algorithm was evaluated by measuring the speech-in-noise performance of 14 CI users using three types of background noise. Two NNSE algorithms were compared: a speaker-dependent algorithm, that was trained on the target speaker used for testing, and a speaker-independent algorithm, that was trained on different speakers. Significant improvements in the intelligibility of speech in stationary and fluctuating noises were found relative to the unprocessed condition for the speaker-dependent algorithm in all noise types and for the speaker-independent algorithm in 2 out of 3 noise types. The NNSE algorithms used noise-specific neural networks that generalized to novel segments of the same noise type and worked over a range of SNRs. The proposed algorithm has the potential to improve the intelligibility of speech in noise for CI users while meeting the requirements of low computational complexity and processing delay for application in CI devices.

Performance Analysis of Adaptive Algorithms for Speech Enhancement Applications

Background/Objectives: The improvement of the speech quality is an essential component in this modern world. Though different methods have been used by many researchers since a long period, still the research field is open to work further. In this paper, we have analyzed the same using adaptive algorithms. Methods/Statistical Analysis: The algorithms like Least Mean Square (LMS) and Recursive Least Square (RLS) are used to improve the quality. Further, State-Space Recursive Least Square (SSRLS) algorithm is developed for enhancement of the noisy speech. The existing Spectral Subtraction (SS) algorithm is verified for two types of noisy speech. Also the standard adaptive algorithms have been experienced for these signals. Findings: Spectral Subtraction method is implemented initially. Then LMS and RLS algorithms are tested for two different types of noisy speech signals. In case of LMS, the improvement of SNR is 8.2822 dB for random noise and 6.5245 dB for babble noise. Whereas for RLS algorithm, the SNR improvement is 11.7152 dB for random noise and 14.3883dB for babble noise. The SNR after enhancement is 43.5409 dB and 47.2456 dB for random noise and for babble noise respectively in case of SSRLS algorithm which is much higher than other algorithms as results suggest. Also the error minimization curve converges more rapidly than the others. So SSRLS algorithm provides better enhanced signal than the other three methods. Application/Improvements: The applications are in different field of communications like mobile telephony, video conferences as audio and video applications. Also the applications are of speech recognition, voice communications, information forensics and many more. The improvement of SNR in this proposed algorithm found better than SS, LMS and RLS algorithms as 31.2377 dB and 35.6381 dB.

Enhanced multiclass SVM with thresholding fusion for speech-based emotion classification

As an essential approach to understanding human interactions, emotion classification is a vital component of behavioral studies as well as being important in the design of context-aware systems. Recent studies have shown that speech contains rich information about emotion, and numerous speech-based emotion classification methods have been proposed. However, the classification performance is still short of what is desired for the algorithms to be used in real systems. We present an emotion classification system using several one-against-all support vector machines with a thresholding fusion mechanism to combine the individual outputs, which provides the functionality to effectively increase the emotion classification accuracy at the expense of rejecting some samples as unclassified. Results show that the proposed system outperforms three state-of-the-art methods and that the thresholding fusion mechanism can effectively improve the emotion classification, which is important for applications that require very high accuracy but do not require that all samples be classified. We evaluate the system performance for several challenging scenarios including speaker-independent tests, tests on noisy speech signals, and tests using non-professional acted recordings, in order to demonstrate the performance of the system and the effectiveness of the thresholding fusion mechanism in real scenarios.

Speech Enhancement using Kalman Filter with Preprocessed Digital Expander in Noisy Environment

Objective: The primary objective of the Speech Enhancement algorithms is to enrich the superiority of speech. The superiority of speech is articulated in two factors, clarity and other is intelligibility. Methods/Statistical Analysis: The method to improve the quality of speech in this paper is proposed based on computationally efficient AR modeled Kalman Filter with digital compressor/expander. This approach is based on reconstruction of noisy speech signal using digital expander and further enhancement with Auto Regressive modeled Kalman filter. Findings: The results of proposed method in terms of SNR and intelligibility are found to be better compared to earlier methods like spectral subtraction; wiener filter and Kalman filter methods. Application/Improvement: This study suggests that improvement in speech signal recovery in noisy environment helping researchers for developing efficient devices in the field of Speech recognition systems, Speech based authentication systems, audio processing devices and so on.

Speech Enhancement Using Iterative Kalman Filter with Time and Frequency Mask in Different Noisy Environment

The main aim of the Speech Enhancement algorithms is to improve the Quality of speech. The Quality of speech is expressed in two parameters. One is clarity, and another is intelligibility. In this paper, we proposed a method to improve the quality of speech based on computationally efficient AR modeled Iterative Kalman Filter with time and frequency mask. This approach is based on reconstruction of noisy speech signals using Auto Regressive modeled Kalman filter and further to reduce artifact noise time and frequency mask is applied to the Kalman filter output. The results of the proposed method are found to be better compared to spectral subtraction, wiener filter and Kalman filter methods.

Mean-Median based Noise Estimation Method using Spectral Subtraction for Speech Enhancement Technique

Background/Objectives: This article proposes a new noise estimation method which is based on mean and median statistical tools. In this article, the proposed method i.e., Mean-Median based noise estimation method have applied in the spectral subtraction method for speech enhancement technique. Methods/Statistical Analysis: The MATLAB/SIMULINK platform is used for simulating the simulation model of proposed technique. For performance evaluation of proposed noise estimation, Perceptual Evaluation of Speech Quality (PESQ) Score and Simulation time are chosen. Different noisy speech signal and clean speech signal have taken as input signal for proposed model and for finding PESQ score, respectively. Findings: The advantage of our proposed noise estimation method, it does not require signal to noise ratio, Voice Activity Detector, or histograms. This paper is carried comparison results with Modified Cascaded Median (MCM) based noise estimation method and also simulation time for different corrupted speech files. Application/Improvements: As per expected, proposed technique is given better speech quality signal (in PESQ scores). As compare to MCM based technique, it takes less simulation time and also no memory storage requirement.

Sparse Representations for Single Channel Speech Enhancement Based on Voiced/Unvoiced Classification

The approach presented here in relies on a new voicing decision algorithm based on the multi-scale product (MP) characteristics. The MP is based on the multiplication of Wavelet Transform Coefficients at some scales. According to the voicing decision, improved subspace decomposition is operated on the voiced segments of the noisy speech signal and a multi-scale principal component analysis is applied on the unvoiced segments of the same signal. Furthermore, the voiced frames are decomposed into three subspaces: sparse, low rank, and the remainder noise components. Then, we calculate the components as a segregation problem. In the unvoiced frames, we combine the straightforward multivariate generalization of the wavelet denoising technique with the principal component analysis method. Experiments on NOIZEUS and NTT databases show that the proposed approach obtains satisfying results for most types of noise with little speech degradation and outperforms several competitive methods.

A wavelet- based transform method for quality improvement in noisy speech patterns of Arabic language

This paper addresses the problem of single-channel speech enhancement of low (negative) SNR of Arabic noisy speech signals. For this aim, a binary mask thresholding function based coiflet5 mother wavelet transform is proposed for Arabic speech enhancement. The effectiveness of binary mask thresholding function based coiflet5 mother wavelet transform is compared with Wiener method, spectral subtraction, log-MMSE, test-PSC and p-mmse in presence of babble, pink, white, f-16 and Volvo car interior noise. The noisy input speech signals are processed at various levels of input SNR range from ź5 to ź25 dB. Performance of the proposed method is evaluated with the help of PESQ, SNR and cepstral distance measure. The results obtained by proposed binary mask thresholding function based coiflet5 wavelet transform method are very encouraging and shows that the proposed method is much helpful in Arabic speech enhancement than other existing methods.

Estimating acoustic speech features in low signal-to-noise ratios using a statistical framework

Accurate estimation of acoustic speech features from noisy speech and from different speakers is an ongoing problem in speech processing. Many methods have been proposed to estimate acoustic features but errors increase as signal-to-noise ratios fall. This work proposes a robust statistical framework to estimate an acoustic speech vector (comprising voicing, fundamental frequency and spectral envelope) from an intermediate feature that is extracted from a noisy time-domain speech signal. The initial approach is accurate in clean conditions but deteriorates in noise and with changing speaker. Adaptation methods are then developed to adjust the acoustic models to the noise conditions and speaker. Evaluations are carried out in stationary and nonstationary noises and at SNRs from −5 dB to clean conditions. Comparison with conventional methods of estimating fundamental frequency, voicing and spectral envelope reveals the proposed framework to have lowest errors in all conditions tested.

EEG-Informed Attended Speaker Extraction From Recorded Speech Mixtures With Application in Neuro-Steered Hearing Prostheses

We aim to extract and denoise the attended speaker in a noisy two-speaker acoustic scenario, relying on microphone array recordings from a binaural hearing aid, which are complemented with electroencephalography (EEG) recordings to infer the speaker of interest. In this study, we propose a modular processing flow that first extracts the two speech envelopes from the microphone recordings, then selects the attended speech envelope based on the EEG, and finally uses this envelope to inform a multichannel speech separation and denoising algorithm. Strong suppression of interfering (unattended) speech and background noise is achieved, while the attended speech is preserved. Furthermore, EEG-based auditory attention detection (AAD) is shown to be robust to the use of noisy speech signals. Our results show that AAD-based speaker extraction from microphone array recordings is feasible and robust, even in noisy acoustic environments, and without access to the clean speech signals to perform EEG-based AAD. Current research on AAD always assumes the availability of the clean speech signals, which limits the applicability in real settings. We have extended this research to detect the attended speaker even when only microphone recordings with noisy speech mixtures are available. This is an enabling ingredient for new brain-computer interfaces and effective filtering schemes in neuro-steered hearing prostheses. Here, we provide a first proof of concept for EEG-informed attended speaker extraction and denoising.

AN EFFICIENT MODIFIED CUCKOO SEARCH BASED NOISE DESTRUCTION AND IMPROVEMENT OF SPEECH SIGNAL QUALITY

The harmonic signal takes its origin from a restricted spectral illustration of a voice signal. It is integrated with at least a section of the restricted delayed spectral signal to furnish a reconstructed speech signal equipped with conceptually enhanced audio quality. Speech signal quality augmentation involves the challenge to scale down the noise and to perk up the transparency. In this document, we valiantly launch an effective noise destruction method for perfection of speech signals by means of the modified cuckoo search algorithm. At first stage, the Input original signal and noisy speech signal features are extracted using the Amplitude Magnitude Spectrogram (AMS). Then the signals are classified in order to generate the initial population. The Second stage is to generate the best weight using modified cuckoo search algorithm and in the third stage, Ranking and Merging is used in order to improve the quality by ranking the best solution from the merging of Modified levy flight and Velocity in PSO. Hence the underlying objective, in this regard, is to boost the speech signal quality and do away with the noise prevalent in the signal. The cheering outcomes achieved have unequivocally ratified effectiveness of our proposed technique and is able to destruct noise and to improve the quality of the speech signal

Speech enhancement using robust estimators and rank-order statistics

Purpose The purpose of the present work is to design robust estimators for speech enhancement by incorporation of calculation rank-order statistics and locally-adaptive neighborhoods. The proposed estimators are able to increase the speech quality of a noisy signal, to preserve better speech intelligibility, and to introduce less artifacts comparing with known speech enhancement estimators. Design/methodology/approach We design a novel speech enhancement algorithm based on rank-order statistics and local adaptive signal processing to improve the accuracy of existing speech enhancement estimators, in terms of speech quality, intelligibility, and introduction of artificial artifacts. Findings We found that by using the proposed estimators for speech enhancement we obtain a better adaptation to nonstationary characteristics of speech and noise processes comparing with that of known speech enhancement estimators. The proposed algorithm increases speech quality, preserves better speech intelligibility, and introduces less artifacts comparing with known speech enhancement estimators. Research limitations/implications The proposed approach for speech enhancement is a locally-adaptive signal processing performed for each element of a noisy speech signal. Thus, the main limitation of the proposed approach is an increase of computational complexity compared with that of nonadaptive conventional techniques. Practical implications In order to perform real-time speech enhancement with the proposed approach, it is recommended to use a digital system with a fast processor. Another option is by using a parallel architecture such as a FPGA. Originality/value We propose a novel local-adaptive algorithm for robust speech enhancement by incorporation of calculation of rank-order statistics and local-adaptive neighborhoods. The proposed algorithm is able to adjust itself in response to changes in the statistical properties of ambience noise.

Speech Enhancement Using Multi‐channel Post‐Filtering with Modified Signal Presence Probability in Reverberant Environment

In this paper, a multi-channel post-filtering approach in reverberant environment based on detection and estimation scheme is presented. A modified Signal presence probability (SPP), which is in consideration of reverberation, is proposed with a novel estimator Direct-toreverberate ratio (DRR) to adapt to distant-talking scene. SPP is known a key estimator to instruct the updating of transient noise or residual directional interference and form gain function in the time-frequency domain, consequently a new desired signal detection scheme is proposed to improve its accuracy. Appropriate spectral enhancement technique is applied to the noisy speech signal taking advantage of the modified SPP estimator. The proposed multi-channel post-filtering is tested in different nonstationary noisy and reverberant environments. Experimental results show that it achieves considerable improvement on signal preservation of the desired speech with more noise reduction over the comparative algorithms.

Speech Enhancement based on Wiener Filter and Compressive Sensing

<p><em>In communication medium to satisfy the speech enhancement process by using differents methodologies and algoirthms are the key term in testing the system design well enough to produce the best performance results for the speech system. The Wiener filter is one of the classical algorithm that applied to speech process to avoid the noise attacking the speech signal. In other word, compressive sensing method by randomize measurement matrix are combined with the Wiener filter to analyse the noisy speech signal with less introduce to noise signal and producing high signal to noise ratio. The PESQ is used to measure the quality of the proposed algorithm design. As in the experimental results shows that, attacking of defferent noise environments in speech signal still effectively improve the performance of noisy speech with maintain the high score of the PESQ quality. </em><em></em></p>

A pre-processing method for improvement of vowel onset point detection under noisy conditions

Vowel onset point (VOP) is the instant of time at which vowel region starts in a speech signal. VOP plays a vital role in different applications of speech processing, such as syllable detection, speaker verification, duration modification, language identification etc. There are different existing algorithms for the detection of instance of VOP in a speech signal. The algorithm based on the combined evidences extracted from the source excitation, spectral peaks and modulation spectrum has been used as a baseline system for the present work. The baseline system performs well under clean speech data. However, under noisy conditions the performance of the baseline system degrades. The performance of the system degrades in terms of more number of spurious VOPs, which get detected under noisy speech conditions. According to the available literature, this degraded performance is due to the spectral broadening of the speech in the noisy environments. In this paper we have proposed a pre-processing technique on top of the baseline system to reduce this spectral broadening effect of noise. The noisy speech data are passed through the pre-processing algorithm in order to minimize the spectral broadening effect of speech signal. The pre-processed speech is then passed through the baseline system to detect the VOPs in the speech signal. Experiments were carried out under clean and different noisy speech signals. The results of the experiment show an improvement of 16–21% in terms of removal of spurious VOPs, over the existing baseline system under different noisy speech conditions. Further, the performance of the proposed method has been compared with two different best performing techniques for detection of VOP, and found that the proposed method gives a superior level of performance in terms of identification accuracy and identification rate.

An efficient frequency-domain adaptive forward BSS algorithm for acoustic noise reduction and speech quality enhancement

In this paper, we consider the speech enhancement and acoustic noise reduction problem in a moving car through a blind source separation scheme employing two loosely spaced microphones. We propose a new efficient frequency domain-symmetric adaptive decorrelation (FD-SAD) algorithm that removes punctual noise components from noisy speech signals. The FD-SAD algorithm is combined with the forward blind source separation FBSS structure to enhance the performances of its time-domain symmetric adaptive decorelating (TD-SAD) version. The proposed algorithm has a good tracking behaviour and fast convergence speed even in very noisy conditions with loosely spaced microphones. Intensive experiments have been done on the newly proposed algorithm in terms of the Segmental Signal-to-Noise-Ratio (SegSNR), the System Mismatch (SM), the Segmental Mean Square Error (SegMSE), and the Cepstral Distance (CD) criteria. The comparison results with the state-of-the-art algorithms have highlighted the excellent performance of the proposed algorithm, and have shown its ability to completely remove the correlated noise components from speech signal even in very noisy conditions when controlled by a voice activity detector.

Phase distortion resulting in a just noticeable difference in the perceived quality of speech

Common speech enhancement methods based on the short-time Fourier analysis–modification–synthesis (AMS) framework, modify the magnitude spectrum while keeping the phase spectrum unchanged. This is justified by an assumption that the phase spectrum can be seen as unimportant to speech quality, and hence the noisy phase spectrum can be used as a reasonable estimate of the clean phase spectrum in signal reconstruction. In this work we show, by using an ideal magnitude estimator, that corruption in the phase spectrum can still affect the quality of the resulting speech in low SNR environments. Furthermore, we quantify the distortion in the phase spectrum which can be tolerated before it begins to affect speech quality. This is done through a series of experiments, using both subjective and objective tests, and statistical analysis to evaluate the results. The results show that the phase spectrum computed from noisy speech can be used as an estimate of the phase spectrum of the clean signal without noticeably affecting perceived speech quality, only if the segmental SNR of the noisy speech signal is greater than 7 dB.